On Wed, May 30, 2012 at 5:39 PM, R. Diez <[email protected]> wrote: > Hallo Pete: > > The l.maci tests fail when built with your toolchain, event though they are > written in assembly. > > I disassembled the generated ELF file, and something is wrong with the > immediate constant for that instruction, look at the differences here: > > 292c: 4d e5 07 ff l.maci r5,0x7ff # Your toolchain > 292c: 4c 05 7f ff l.maci r5,0x7fff # The old toolchain > > 2a98: 4f e5 07 fd l.maci r5,0x7fd # Your toolchain > 2a98: 4c 05 ff fd l.maci r5,0xfffffffd # The old toolchain > > 2dd8: 4e 05 00 00 l.maci r5,0x0 # Your toolchain > 2dd8: 4c 05 80 00 l.maci r5,0xffff8000 # The old toolchain
Yeah, I had some problems with that before, too. I think maybe I used the description of the encoding from the architecture manual, which isn't actually correct. I'll take a look at it. By the way, I've been working on integrating your simulator testsuite with the or1k-src tree so you can just use "make check" to test it. I've found quite a few bugs that way :) -Pete _______________________________________________ OpenRISC mailing list [email protected] http://lists.openrisc.net/listinfo/openrisc
